Voice AI at scale

Voice Infra runs voice agents that call your customers, answer your business number, and talk to people in the browser and on WhatsApp. The same API key also gives you the speech models underneath: text to speech, speech to text and a language model, each with its own API. Agents speak and understand Hindi, Gujarati, Marathi, Bengali, Tamil, Telugu, Kannada and Malayalam, and the Hinglish that most real calls are in.

Start here

Quickstart

Create a key and an agent, then place one real phone call in five minutes.

Start building

For partners: your first campaign, the report and the go-live checklist.

API overview

One auth header, money in rupees, pagination and idempotency keys.

API reference

Every endpoint, with a request you can send from the page and a Postman collection.

Voice agents

Agents

The prompt, opening line, voice and tools a call runs with. Single-prompt or node-based.

Phone calls

Place outbound calls and read each transcript and recording.

Campaigns

Dial a contact list with calling hours, retries, a budget and a report.

Phone numbers

Pool your numbers, set how fast each one calls and watch its health.

Inbound calls

Answer your business number with an agent.

Browser calls

Run an agent in your user’s browser, with live captions.

WhatsApp calls

Voice calls on the WhatsApp numbers your brands already use.

Call results

A summary, an outcome and the fields you asked for, after every call.

Webhooks

Signed events as each call starts and ends, with the outcome.

Sub-workspaces

Run many brands from one live workspace and one key.

Tools

Call your own APIs before, during and after a call.

Knowledge bases

Give an agent your documents to answer from.

Speech APIs

Text to speech

One HTTP request for a clip, or a WebSocket for live audio. OpenAI-compatible.

Speech to text

Transcribe a clip or stream live audio, in ten Indian languages.

Recorded calls

Turn a call recording into a transcript with speakers and timestamps.

Language model

mira-llm on OpenAI’s chat completions, with streaming and tool calls.

Pipecat

Use our speech models inside the Pipecat bot you already run.

LiveKit

Use our voices in a LiveKit agent.

Pricing

Agent minutes cost ₹5 pay as you go, ₹3 with a ₹50,000 monthly commitment, or ₹1 with a ₹1,00,000 monthly commitment. Every company starts on pay as you go. On the ₹1 plan, call insights (₹0.20 a minute, on by default) and our phone line (₹0.30 a minute, if you use it) are added on top. Prices exclude GST. Billing & tiers explains each plan and how a call is billed. The speech APIs are priced on their own; see the Speech API and language model pages.

Your first call

Open the console, go to Developers and press Create key. The first time, the console asks for your work email and sends you a confirmation link. The key is shown once, so put it in your secret manager. Create an agent:

curl -X POST https://sandbox.voice.miraiminds.co/v2/agents \
  -H "Authorization: Bearer sk_live_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Order confirmation",
    "system_prompt": "You are Ashu from Acme. Confirm the order and ask whether the delivery address is unchanged. Keep replies short. When the customer is done, thank them and end the call.",
    "first_message": "नमस्ते, मैं Acme से आशु बोल रहा हूँ।",
    "voice": { "voice_id": "ashu", "language": "hi-IN" },
    "language": "hi-IN"
  }'

Then call a phone you are allowed to call, with the id you got back:

curl -X POST https://sandbox.voice.miraiminds.co/v2/calls \
  -H "Authorization: Bearer sk_live_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "agent_id": "agt_01JZQ8F3K7M2N5P9R4T6V8W0XZ", "to": "+919876543210" }'

The Quickstart has both requests in Python and Node, and the webhook that tells you how the call went.

Start here

Build

Prompting

Speech API

Text to speech

Speech to text

Language model

Money

Reference

Start here

Quickstart

Start building

API overview

API reference

Voice agents

Agents

Phone calls

Campaigns

Phone numbers

Inbound calls

Browser calls

WhatsApp calls

Call results

Webhooks

Sub-workspaces

Tools

Knowledge bases

Speech APIs

Text to speech

Speech to text

Recorded calls

Language model

Pipecat

LiveKit

Pricing

Your first call

​Start here

Quickstart

Start building

API overview

API reference

​Voice agents

Agents

Phone calls

Campaigns

Phone numbers

Inbound calls

Browser calls

WhatsApp calls

Call results

Webhooks

Sub-workspaces

Tools

Knowledge bases

​Speech APIs

Text to speech

Speech to text

Recorded calls

Language model

Pipecat

LiveKit

​Pricing

​Your first call

Start here

Voice agents

Speech APIs

Pricing

Your first call